A method and system for highway traffic incident active monitoring and emergency disposal

Abstract

The application discloses a highway traffic event active monitoring and emergency disposal method and system, designs a five-layer system architecture, collects and fuses data generated by roadside equipment, vehicle-mounted terminals, third-party internet platforms and the like, carries out data feature extraction and multi-modal fusion, realizes event active detection and automatic review, realizes event disposal decision dynamically according to traffic event data, emergency resource conditions and road network topological structures, and pushes road condition related instruction information to travel service terminals.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of expressway traffic safety management, and in particular to an expressway traffic event active monitoring and emergency disposal method and system. BACKGROUND

[0002] With the continuous growth of expressway traffic density, traffic accidents, abnormal parking, and sudden events induced by bad weather occur frequently. The traditional disposal mode relying on manual patrol and passive alarm has been difficult to meet the modern traffic management needs. Realizing the "second-level discovery, minute-level disposal" of traffic events has become the core goal of the industry. The technical key lies in breaking through the bottlenecks of insufficient all-time domain perception accuracy, low multi-system coordination efficiency, and unscientific disposal resource scheduling.

[0003] The existing expressway event discovery mainly relies on manual patrol, rescue calls, or single sensors such as video monitoring, which requires manual monitoring. The event discovery efficiency is low, the coverage is limited, and the event discovery is not timely, which delays the disposal golden time. The current industry is experiencing a technical transformation from "passive response" to "active prevention and control". With the development of intelligent expressway construction and intelligent transportation technology, roadside devices such as millimeter wave radars and weather sensors, vehicle terminal systems such as GPS and Internet of Vehicles platforms, and third-party Internet platforms such as map navigation and social media have the function of sudden event discovery and reporting. However, due to the dispersion of data, there is a lack of effective fusion and analysis, making it difficult to form a comprehensive and accurate event situation awareness. In addition, when organizing emergency event disposal, the scheduling of emergency resources such as rescue vehicles and personnel is often based on experience, lacking scientific basis and dynamic optimization, which may lead to idle or unscientific resources.

[0004] Although there are some event detection algorithms and emergency disposal platforms at present, the main shortcomings of the existing technology and platforms are that the perception data acquisition ability and decision-making intelligence level are seriously mismatched, and manual decision-making leads to disposal delay. The specific performance is as follows: relying on a single data source, the accuracy and robustness are insufficient. After detecting the event, a large amount of manual intervention is still needed for confirmation, information flow, and command and dispatch. The monitoring, confirmation, reporting, decision-making, scheduling, execution, and feedback are fragmented, and do not form a closed loop. There is a lack of intelligent decision support, and the disposal scheme is generated slowly and is not optimized. SUMMARY

[0005] The present application aims to overcome the shortcomings of the prior art and proposes an expressway traffic event active monitoring and emergency disposal method and system. The core principle is to replace the decision-making layer with machine intelligence. Through multi-source fusion, the reliability of decision-making input is ensured. A cross-platform collaboration is established to open up the information channel. A dynamic optimization model is used to improve resource efficiency. The disposal response is compressed within the golden time window. The improvement from "human brain decision-making lag" to "system intelligent second-level response" is realized. The core pain points of the low efficiency of expressway safety management disposal are solved.

[0006] The purpose of the present application is achieved by the following technical solutions: a highway traffic event active monitoring and emergency disposal method, comprising:

[0007] S1, collect multi-source traffic event original data and perform feature extraction and space-time alignment, construct a multi-modal fusion model and perform training, input the obtained fusion features into the trained multi-modal fusion model for AI detection, and output traffic event elements;

[0008] S2, establish an emergency resource scheduling model based on reinforcement learning, the emergency resource scheduling model inputs traffic event elements, emergency resource states and road network topology, performs similar historical case retrieval on the traffic event elements, extracts a matched basic disposal framework, and performs resource scheduling optimization, while combining real-time information for dynamic correction, and outputs an optimal scheme;

[0009] S3, push instructions to terminals through a standardized API protocol;

[0010] S4, generate an accident report through a large language model after the event disposal is completed.

[0011] Further, the multi-source traffic event original data includes: road network data, roadside device data, vehicle-mounted terminal data, and third-party platform data.

[0012] Further, the space-time alignment includes: using global navigation satellite system timestamps + milepost numbers to match and unify different source data to a space-time coordinate system for space-time alignment; time alignment: taking the Beidou satellite timing system as a reference, correcting the time stamps of each device; space alignment: converting radar polar coordinates into GIS longitude and latitude, and matching with video detection frames.

[0013] Further, the multi-modal fusion model includes: a CNN branch, a Transformer branch, and a multi-modal fusion layer,

[0014] The CNN branch is used to process data with spatial structure;

[0015] The Transformer branch is used to process time series data and global information, and uses attention mechanisms to capture the correlation between different modalities, positions and time points;

[0016] The multi-modal fusion layer is used to integrate the local spatial features extracted by the CNN branch and the global space-time features extracted by the Transformer branch;

[0017] Further, the training process of the multi-modal fusion model includes two-stage training: in the first stage, the pre-training weights of the CNN and the Transformer are frozen, only the fusion module and the task-specific decoder are trained; in the second stage, all weights are unfrozen, end-to-end joint fine-tuning is performed, all components work together to achieve global optimization; the loss function used in training is obtained by weighting the classification loss, position regression loss and influence range loss.

[0018] Further, the emergency resource scheduling model includes: first, pre-plan matching is performed, similar historical cases are retrieved from the knowledge base, the basic treatment framework is extracted, and a mathematical optimization model considering event characteristics, emergency resources and road network state is established; model input and parameter definition are constructed, optimization objective and constraint condition are set, and a mixed integer programming solver (MIP Solver) combined with genetic algorithm is used for solving; the rescue vehicle arrival time is calculated in real time, and if the timeout exceeds the set threshold, the standby vehicle scheduling is started, and the traffic control measures, resource scheduling instructions and public guidance information are output.

[0019] Further, the instruction pushing to the terminal through the standardized API protocol includes: using structured JSON data packets when pushing instructions to public travel terminals, including fields: event ID, instruction type, location and execution terminal; IPv6 addresses are allocated for each type of terminal, and the transmission is performed through the 5G URLLC channel;

[0020] The traffic control department system, the navigation platform and the vehicle terminal distribute instructions according to corresponding protocols.

[0021] According to another aspect of the specification, a highway traffic event active monitoring and emergency disposal system is also provided, which includes:

[0022] The perception layer, the network layer, the data layer, the intelligent hub layer and the application layer;

[0023] The perception layer is used to obtain multi-source traffic event raw data,

[0024] The network layer is used to provide communication transmission;

[0025] The data layer is used to save real-time databases, fusion feature databases, resource state databases and pre-plan knowledge bases; the real-time database includes real-time updated data, the fusion feature database includes multi-dimensional vectors after time and space alignment, the resource state database includes rescue vehicle real-time position and available resource type data; and the pre-plan knowledge base includes disposal pre-plans;

[0026] The intelligent core layer includes an AI event detection engine, a dynamic decision engine, and an instruction distribution engine; the AI event detection engine aligns multi-source traffic event raw data in space and time, processes multi-source fusion data using a multi-modal fusion model, and outputs result data; the dynamic decision engine is used for pre-plan matching, resource scheduling optimization, and real-time dynamic correction based on reinforcement learning, and generates an optimal disposal scheme; and the instruction distribution engine decomposes the scheme into executable instructions through a standardized API protocol.

[0027] The application layer is used for realizing human-computer interaction and instruction execution.

[0028] According to another aspect of the specification, a highway traffic event active monitoring and emergency disposal device is also provided, including a memory and one or more processors, the memory stores executable code, and the processor executes the executable code to realize the highway traffic event active monitoring and emergency disposal method.

[0029] According to another aspect of the specification, a computer readable storage medium is also provided, which stores a program, and the program is executed by a processor to realize the highway traffic event active monitoring and emergency disposal method.

[0030] The beneficial effects of the present application are:

[0031] (1) Full-process AI autonomous closed loop: the response time of the whole link of "monitoring-decision-execution-feedback" is compressed to within 2 minutes;

[0032] (2) Multi-modal fusion detection: millimeter wave radar + video + floating car data space-time alignment fusion, which greatly improves the accuracy and effectiveness of event identification;

[0033] (3) Dynamic optimization decision: resource scheduling model based on reinforcement learning, reducing the empty running rate of rescue vehicles to 8%;

[0034] (4) Cross-platform instruction coordination: standardized protocol stack connects multi-source heterogeneous systems, reducing instruction synchronization delay. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 A highway traffic event active monitoring and emergency disposal method flowchart is provided for the embodiments of the present application;

[0036] Figure 2 A multi-source data acquisition and fusion analysis flowchart is provided for the embodiments of the present application;

[0037] Figure 3 A highway traffic event active monitoring and emergency disposal system architecture diagram is provided for the embodiments of the present application;

[0038] Figure 4 A schematic diagram of a highway traffic event active monitoring and emergency disposal device provided for an embodiment of the present application. DETAILED DESCRIPTION

[0039] The specific embodiments of the present application are further described in detail below with reference to the accompanying drawings.

[0040] As Figure 1 shown, the present application provides a highway traffic event active monitoring and emergency disposal method and system, which designs a five-layer system architecture, collects and fuses data generated by roadside devices, vehicle terminals, third-party Internet platforms, etc., performs data feature extraction and multi-modal fusion, realizes event active detection and automatic review; According to the traffic event data, emergency resource conditions and road network topology, dynamically realize event disposal decision, and push the road condition related instruction information to the travel service terminal; At the same time, a dynamic correction mechanism is designed, which follows up the event disposal situation in real time, and automatically triggers the adjustment scheme for the steps not executed as expected.

[0041] As Figure 1 shown, a highway traffic event active monitoring and emergency disposal method specifically includes the following steps:

[0042] Step 1: Multi-source data collection and fusion analysis, as Figure 2 shown, including:

[0043] (1) Real-time collection of multi-source data: Obtain raw data from roadside devices, vehicle terminals, third-party platforms, etc., and perform feature extraction.

[0044] (2) Time and space alignment: Use GNSS (Global Navigation Satellite System) timestamp + mileage marker matching to unify different source data to the time and space coordinate system (error <0.1s / ±0.5m) for time and space alignment. Time alignment: Use Beidou satellite timing system as the reference to correct the time stamp of each device; Space alignment: Convert radar polar coordinates to GIS (Geographic Information System) latitude and longitude, and match with video detection box.

[0045] (3) AI event detection: Use a multi-modal fusion model (CNN+Transformer structure) to input fused features and output event five elements: type (accident / congestion / throwing object, etc.), location (marker + lane), time, confidence, and impact range.

[0046] Model architecture design: Considering that the data input from roadside devices, vehicle terminals, third-party platforms, etc. is multi-source heterogeneous data, highway traffic events have spatio-temporal correlation, and abnormal events such as traffic accidents, vehicle breakdowns, and congestion have different characteristics, this scheme designs a "double branch-fusion-decoding" model architecture.

[0047] (1) CNN branch: Specialized in handling data with spatial structure, such as video images and radar point clouds. It is responsible for extracting visual features of targets (such as vehicles, pedestrians, and foreign objects) and spatial scene features (such as lane lines and road surfaces).

[0048] (2) Transformer branch: Specialized in handling time series data (such as trajectories and traffic flow) and global information, and utilizing its powerful "Attention" mechanism to capture complex correlations between different modalities, different locations, and different time points.

[0049] (3) Multi-modal fusion layer: This is the core of the model, responsible for efficiently integrating local spatial features extracted by CNN and global spatio-temporal features extracted by Transformer, forming a unified and comprehensive environmental perception representation.

[0050] (4) Task-specific decoder: Based on the unified representation after fusion, it outputs five elements such as event type, location, and confidence.

[0051] Model training process design:

[0052] (1) Training data preprocessing. First, highway target detection dataset construction. From highway scene videos, 10000 images are selected for dataset construction, with a video frame rate of 25 FPS, saving 1 frame every 100 frames, resulting in a total of 52144 saved images, covering all detection scenarios. Using manual annotation, each event sample is accurately labeled with type, post number, lane, and impact length. Second, spatial and temporal enhancement of training data. Randomly rotate image frames (small angle), adjust brightness, change contrast, simulate rain and fog weather noise, etc. to achieve spatial enhancement; randomly sample video clips to simulate different event occurrence speeds, achieving temporal enhancement. Finally, standardize radar data and vehicle trajectory data using Z-score standardization method.

[0053] (2) Loss function design. Considering that a traffic scene may contain multiple event types at the same time, such as a vehicle collision accident that may cause traffic congestion and road debris, event type detection is essentially a multi-label classification problem. To effectively train the model for multi-label classification, the loss function is designed as the weighted sum of multiple parts, as follows:

[0054] Formula explanation:

[0055] Classification loss (L cls): Focal Loss. There are much more normal driving scenes than abnormal events on highways. Using Focal Loss can automatically reduce the focus on a large number of "easy samples" (i.e. normal scenes that are easy to classify), and focus on a small number of "difficult samples" (i.e. various abnormal events), thereby effectively improving the model's detection ability for rare events.

[0056] Position regression loss (L loc): Smooth L1 Loss. Calculate the error between the predicted and true milepost, predicted and true lane, respectively.

[0057] Impact range loss (L impact): Smooth L1 Loss, similar to position regression, used to calculate the prediction error of impact length and lane number.

[0058] Weight coefficients : determined by tuning on the validation set to balance the importance of different tasks.

[0059] (3) Optimizer: use AdamW optimizer to decouple weight decay from gradient updates, to prevent model overfitting.

[0060] (4) Two-stage training:

[0061] First stage: freeze the pre-training weights of CNN and Transformer, only train the fusion module and task-specific decoder. This allows the newly added parts to quickly find a reasonable solution space.

[0062] Second stage: unfreeze all weights, perform end-to-end joint fine-tuning, so that all components work together to achieve global optimization.

[0063] Step two: dynamic situation assessment and decision-making.

[0064] Based on the traffic events actively detected by AI in step one, the emergency response process is triggered in the first time. Specifically as follows:

[0065] (1) Pre-plan matching: retrieve similar historical cases (matching degree > 85%) from the knowledge base, and extract the basic disposal framework (such as "accident needs to close the lane + dispatch rescue").

[0066] (2) Resource scheduling optimization: establish a mathematical optimization model that considers event characteristics, emergency resources, and road network status, to automatically and intelligently generate the optimal scheduling scheme, as follows:

[0067] First step: model input and parameter definition. One is event information from AI detection model, including: event type (such as traffic accident, fire, serious congestion, cargo scattering, etc.), event location (stake information, up / down driving direction), event level (determined by AI model according to influence range, vehicle queue length, etc. Level, divided into I level: very serious, II level: serious, III level: general), predicted influence range (predicted event disposal time, congestion spread range and duration). Two is emergency resource information, including: resource set (rescue center, obstacle removal point and its contained obstacle removal vehicle, ambulance, fire truck, police car, emergency personnel, etc.), resource state (idle, on the way, busy), current location; Resource capacity (each resource unit can handle event type). Three is road network basic data and state information, road network basic data mainly includes road network topology, road network state information is updated by traffic flow data The real-time traffic time of each road section.

[0068] Second step: optimization target and constraint condition. Optimization target includes: one is the shortest total response time, the total time of all required resources reaching the scene from the instruction is the smallest, which is the highest priority target; Two is the lowest total scheduling cost, considering the fuel consumption, wear and tear, labor cost, etc. of the dispatched vehicles; Three is the maximum disposal effect, preferentially dispatching professional resources with the most matched disposal capacity and event level and the highest efficiency; Constraint conditions include: one is the demand satisfaction constraint, the dispatched resource combination must completely satisfy all resource types and quantities required for disposing the event of this type and level, such as 2 accidents require at least 2 obstacle removal vehicles; Two is the resource uniqueness constraint, the same resource unit can only be dispatched to one event point at the same time; Three is the time window constraint, the resource must arrive within the required time window; Four is the path feasibility constraint, the calculated path must be actually passable, considering traffic congestion and lane closure, etc.

[0069] Third step: model solution and output. Solution algorithm: a combination of mixed integer programming solver (MIP Solver) and genetic algorithm is used for solution. For small-scale problems (few events and resources), MIP is used to obtain accurate solution; For large-scale and real-time requirement high scenes, improved genetic algorithm is used for efficient approximate solution, and the solution is given within seconds. The model outputs the optimal scheduling scheme, which clearly shows "which resources to send", "from which point to send", "which route to take" to "which event point".

[0070] (3) Dynamic correction: real-time receive rescue vehicle ETA (estimated time of arrival), provide ETA for each dispatched resource. When a new event occurs or the original event is upgraded, the model is re-planned, and the travel route or target point of the existing resource is dynamically adjusted, and the optimal scheme containing three elements is output:

[0071] (1) Traffic control measures, such as closing the first and second lanes.

[0072] (2) Resource dispatch instructions, such as: dispatching rescue vehicle A, which is 3km away from the incident site.

[0073] (3) Public guidance information, such as: There is an accident 2km ahead, it is recommended to detour via the XX Expressway exit.

[0074] Step 3: Intelligent distribution of instructions. Instructions are pushed to public travel terminals via standardized API protocols.

[0075] (1) Standardized API: Designed as a structured JSON data packet, containing fields (json):

[0076] {

[0077] Event ID: A20240813001

[0078] "Instruction Type": "Lane Closure"

[0079] Location: G65 Expressway K153+200 Second Lane

[0080] "Execution Terminal": "VMS Device ID007, XX Map"

[0081] }

[0082] (2) When sending instructions to the traffic police system, navigation platform, and vehicle terminal, the following protocols must be followed:

[0083] Traffic police system: Compliant with GA / T 1476-2018 "Traffic Management Data Exchange Protocol"

[0084] Navigation platform: Adapted to navigation map SDK (Software Development Kit)

[0085] Vehicle-mounted terminal: Supports JT / T 808-2019 "Road Transport Vehicle Satellite Positioning Protocol"

[0086] (3) Public travel terminals include four types: roadside VMS (variable information signs), traffic police and rescue personnel APP, vehicle navigation, and Internet map service provider backend.

[0087] Step 4: Automatic Incident Reporting. Once the incident has been handled, an incident report is automatically generated using technologies such as the Big Prophecy Model, completing the incident review.

[0088] like Figure 3 As shown, corresponding to the aforementioned embodiment of a method for proactive monitoring and emergency response to highway traffic incidents, the present invention also provides a system for proactive monitoring and emergency response to highway traffic incidents, the system comprising:

[0089] a perception layer, a network layer, a data layer, an intelligent hub layer, and an application layer;

[0090] The perception layer is configured to acquire multi-source traffic event raw data, including road network data (basic data, traffic flow, and emergencies), roadside devices (cameras, millimeter wave radars, and weather stations), vehicle terminals (OBU: on-board unit, on-board GPS), and third-party platforms (Internet map providers and social media), and collect multi-source heterogeneous data such as videos, radar point clouds, vehicle GPS, and weather.

[0091] The network layer is configured to provide communication transmission; including a 5G / V2X (vehicle-road cooperation) communication network and a fiber transmission network, and the transmission is ensured through a 5G+V2X double-channel transmission, so that the instruction issuing delay is less than 100 ms.

[0092] The data layer is configured to save a real-time database, a fusion feature library, a resource state library, and a preplan knowledge base; the real-time database includes real-time updated data, the fusion feature library includes multi-dimensional vectors after time-space alignment, the resource state library includes real-time position and availability resource data of rescue vehicles, and the preplan knowledge base includes disposal preplans.

[0093] The intelligent hub layer includes an AI event detection engine, a dynamic decision engine, and an instruction distribution engine; the AI event detection engine aligns multi-source traffic event raw data in time and space, processes multi-source fusion data by using a multi-modal fusion model, and outputs result data; the dynamic decision engine is configured to perform preplan matching, resource scheduling optimization, and real-time dynamic correction based on reinforcement learning, and generate an optimal disposal scheme; and the instruction distribution engine decomposes the scheme into executable instructions through a standardized API protocol.

[0094] The application layer is configured to realize human-computer interaction and instruction execution, including a command center large screen, a rescue vehicle terminal, and a public information publishing platform, to realize human-computer interaction and instruction execution, and support command center monitoring, rescue vehicle navigation, and public information synchronous pushing.

[0095] Corresponding to the foregoing embodiment of the highway traffic event active monitoring and emergency disposal method, the present application also provides an embodiment of a highway traffic event active monitoring and emergency disposal device.

[0096] Referring to Figure 4 The embodiment of the present application provides a highway traffic event active monitoring and emergency disposal device, which includes a memory and one or more processors, the memory stores executable code, and the processor executes the executable code to realize the highway traffic event active monitoring and emergency disposal method in the foregoing embodiment.

[0097] The embodiment of the expressway traffic event active monitoring and emergency disposal device provided by the application can be applied to any device with data processing capability, which can be a device such as a computer or the like. The device embodiment can be realized by software, or realized by hardware or a combination of software and hardware. Taking software realization as an example, as a logically meaningful device, it is formed by reading the corresponding computer program instructions in the non-volatile memory into the memory and running by the processor of the device with data processing capability. Figure 4 As shown in the figure, it is a hardware structure diagram of the device with data processing capability of the expressway traffic event active monitoring and emergency disposal device provided by the application, in addition to the processor, the memory, the network interface and the non-volatile memory shown in the figure, the device with data processing capability in the embodiment usually includes other hardware according to the actual function of the device with data processing capability, and details are not described here. Figure 4

[0098] The implementation process of the functions and roles of each unit in the above device is specifically described in the implementation process of the corresponding steps in the above method, and details are not described here.

[0099] For the device embodiment, since it basically corresponds to the method embodiment, the related part can be referred to the part of the method embodiment. The above described device embodiment is only schematic, and the units described as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, that is, they can be located in one place, or distributed on multiple network units. According to the actual needs, part or all of the modules can be selected to realize the purpose of the application scheme. Those skilled in the art can understand and implement without creative labor.

[0100] The embodiment of the application further provides a computer readable storage medium, which stores a program, and the program is executed by a processor to realize the expressway traffic event active monitoring and emergency disposal method in the above embodiment.

[0101] ​The computer readable storage medium can be an internal storage unit of any of the aforementioned data processing capable devices, such as a hard disk or a memory. The computer readable storage medium can also be an external storage device of any of the aforementioned data processing capable devices, such as a plug-in hard disk, a Smart Media Card (SMC), an SD card, a Flash Card, etc. Further, the computer readable storage medium can include both an internal storage unit and an external storage device of any of the aforementioned data processing capable devices. The computer readable storage medium is used to store the computer program and other programs and data required by the aforementioned data processing capable devices, and can also be used to temporarily store data that has been output or will be output.

[0102] The application further provides a computer program product comprising a computer program, which, when executed by a processor, implements the expressway traffic event active monitoring and emergency disposal method.

[0103] Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. It is intended that the specification and examples be considered as exemplary only, with the true scope and spirit of the application being indicated by the following claims.

[0104] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application. The application is not restricted to the precise construction and combinations of parts and steps described above and shown in the accompanying drawings, as such variations and modifications are intended to be included within the scope of the application. The scope of the application is limited only by the claims that follow.

Claims

1. A method for proactive monitoring and emergency response to highway traffic incidents, characterized in that, The method includes: S1. Collect raw data of multi-source traffic events, extract features, and perform spatiotemporal alignment. Construct a multimodal fusion model and train it. Input the obtained fused features into the trained multimodal fusion model for AI detection and output traffic event elements. The multimodal fusion model includes: a CNN branch, a Transformer branch, and a multimodal fusion layer. The CNN branch is used to process data with spatial structure; The Transformer branch is used to process time-series data and global information, and uses an attention mechanism to capture the correlation between different modalities, locations and time points; the multimodal fusion layer is used to integrate the local spatial features extracted by the CNN branch with the global spatiotemporal features extracted by the Transformer branch; The training process of the multimodal fusion model includes two stages: Phase 1: Freeze the pre-trained weights of CNN and Transformer, and train only the fusion module and task-specific decoder; The second stage involves unfreezing all weights and performing end-to-end joint fine-tuning to enable all components to work together and achieve global optimality. The loss function used for training is obtained by weighting classification loss, location regression loss, and influence range loss. S2. An emergency resource scheduling model is established based on reinforcement learning. The emergency resource scheduling model takes into account traffic event elements, emergency resource status and road network topology, performs similar historical case retrieval on traffic event elements, extracts matching basic handling frameworks, optimizes resource scheduling, and dynamically corrects the solution by combining real-time information. S3. Push instructions to the terminal via standardized API protocols; S4. After the incident is handled, an incident report is generated using a large language model.

2. The method for active monitoring and emergency response to highway traffic incidents according to claim 1, characterized in that, The original data of the multi-source traffic incidents includes: road network data, roadside equipment data, vehicle terminal data, and third-party platform data.

3. The method for active monitoring and emergency response to highway traffic incidents according to claim 1, characterized in that, The spatiotemporal alignment includes: using the Global Navigation Satellite System timestamp + mileage marker matching to unify different source data into a spatiotemporal coordinate system for spatiotemporal alignment; time alignment: using the BeiDou satellite timing system as a reference to correct the timestamps of each device; spatial alignment: converting radar polar coordinates into GIS latitude and longitude and matching them with video detection frames.

4. The method for active monitoring and emergency response to highway traffic incidents according to claim 1, characterized in that, The emergency resource dispatch model includes: first, performing contingency plan matching, retrieving similar historical cases from the knowledge base, extracting the basic handling framework, constructing a multi-objective function to minimize response time, congestion losses, and the risk of empty runs by rescue vehicles, and setting constraints including the load capacity of the clearing vehicle and the handling time; using the Q-learning algorithm to solve the optimal dispatch scheme, and receiving the estimated arrival time of the rescue vehicle in real time. If the timeout exceeds a set threshold, the dispatch of backup vehicles is activated, and traffic control measures, resource dispatch instructions, and public guidance information are output.

5. The method for active monitoring and emergency response to highway traffic incidents according to claim 1, characterized in that, The method of pushing instructions to terminals via a standardized API protocol includes: using structured JSON data packets when pushing instructions to public travel terminals, including fields such as: event ID, instruction type, location, and executing terminal; assigning an IPv6 address to each type of terminal and transmitting the data through a 5G URLLC channel; Instructions are distributed to traffic control systems, navigation platforms, and vehicle terminals according to the corresponding protocols.

6. A highway traffic incident active monitoring and emergency response system for implementing the method of any one of claims 1-5, characterized in that, The system includes: The layers are: perception layer, network layer, data layer, intelligent hub layer, and application layer. The perception layer is used to acquire raw data of multi-source traffic events. The network layer is used to provide communication transmission; The data layer is used to store a real-time database, a fusion feature library, a resource status library, and a contingency plan knowledge base. The real-time database includes data that is updated in real time. The fusion feature library includes spatiotemporally aligned multidimensional vectors. The resource status library includes real-time location and availability resource data of rescue vehicles. The contingency plan knowledge base includes emergency response plans. The intelligent central layer includes an AI event detection engine, a dynamic decision engine, and an instruction distribution engine. The AI ​​event detection engine aligns the original data of multi-source traffic events in time and space, processes the multi-source fusion data using a multi-modal fusion model, and outputs the result data. The dynamic decision engine is used to perform contingency plan matching, resource scheduling optimization, and real-time dynamic correction based on reinforcement learning to generate the optimal handling plan. The instruction distribution engine decomposes the plan into executable instructions through a standardized API protocol. The application layer is used to implement human-computer interaction and command execution.

7. A device for active monitoring and emergency response to highway traffic incidents, comprising a memory and one or more processors, wherein the memory stores executable code, characterized in that, When the processor executes the executable code, it implements a method for proactive monitoring and emergency response to highway traffic incidents as described in any one of claims 1-5.

8. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements a method for proactive monitoring and emergency response to highway traffic incidents as described in any one of claims 1-5.

Images